 Why Alternative Vote/Instant Runoff Vote?
When people vote, they often have more than one objective. The most common objective when people vote are the following:
- Ensuring that one's favourite candidate wins
- Ensuring that one's least favourite candidate loses
- Ensuring that the winning candidate is one of one's favourites
- Honestly "answering" the question put to one on the ballot
In many voting systems, one or more of these objectives come into conflict. Depending on which of the above four priorities the voter places first, his ballot will look very different. There are lots of examples of these problems in real world politics. When a voter chooses one of these objectives over the others, people label this behaviour as "strategic voting."
While the Ennies are not as consequential as voting in a real world election, strategic voting can be just as big a problem as in a national election. The reason for this is simple: gamers, especially those on EN World, have to master systems more complex than simple voting systems. As a result, they can easily distinguish between the above four objectives, figure out which is more important, and design a voting strategy accordingly.
How, then, do we solve the problem of strategic voting for the Ennies? By choosing a system in which the above four impulses are not in conflict. Instant runoff voting/alternative vote is a system that has been used in many countries for local and national elections for just this purpose. Good web resources include:
Wikipedia: Instant-runoff Voting
Fairvote Minnessota: Instant Runoff Voting Primer
Fairvote: IRV America
Sample Problems
Rating and Borda Count Systems
Voting systems in which the voter must rate all the products, both those he likes and those he dislikes are notorious for producing strategic voting. The most famous example of these systems are Borda count systems in which every voter is told to rank all his choices in order from favourite to least favourite; points are then assigned in reverse order (a product with a rank of 1 receives 5 points; 2 receives 4; 3 receives 3; 4 receives 2; 5 receives 1).
A sample Borda system ballot might look like this:
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | 1 |
| Sociobiological Roots of Gaming, Ron Edwards | 5 |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | 4 |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 3 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | 2 |
Let's suppose that the voter, in this example, is a fan of mainstream D&D and actually dislikes products that make it and its designers look bad. However, let us suppose that the Gygax book and the Ranger book are practically neck and neck, with the Gygax book lagging just slightly behind. Now, the voter has divided loyalties. Yes, it is important to ensure that the Parker and Edwards books are defeated; but is this more or less important than ensuring that the Gygax book comes out on top? If ensuring Gygax wins is more important to the voter than guaranteeing the defeat of the Edwards book, then it is in his interest to mark the ballot as follows:
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | 1 |
| Sociobiological Roots of Gaming, Ron Edwards | 4 |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | 3 |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 2 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | 5 |
On the first ballot, in which the voter accurately represented his preferences, Gygax gains 1 point on Ranger; but on the second, in which the voter voted strategically, Gygax gains 4 on Ranger. Rating systems, such as those used by EN World last year, encourage voters to review products inaccurately. And often, the better the product is, the greater the incentive to be inaccurate.
Here we see that priority #1 (ensuring that one's favourite product wins) beats out priority #4 (answering the ballot question accurately).
Plurality Systems
In plurality systems, voters can mark an 'x' for products they consider to be worthy of their support. Let's assume that the same products are in the running but the Edwards book is doing very well, running practically neck and neck with the Ranger book. Although the voter likes the Gygax book better than the Ranger book, the Gygax book is lagging behind. So, while this ballot represents his preferences most accurately,
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | X |
| Sociobiological Roots of Gaming, Ron Edwards | |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | |
it is more important that Edwards be defeated than that his favourite wins. So he votes like this:
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | |
| Sociobiological Roots of Gaming, Ron Edwards | |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | X |
Here, we see that priority #2 beats out priorities #1 and #4.
How Does One Mark an AV/IRV Ballot?
AV/IRV ballots are marked very similarly to ballots counted under the Borda system. In Borda systems, as shown above, the voter numbers all candidates in order from favourite to least favourite. Under a Borda system, he would give the two lowest ranks (4 and 5) to the anti-D&D products. In an AV/IRV system, the voter withholds his vote from any product he does not want to see win under any circumstances. Thus, the same ballot marked under this system would look like this:
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | 1 |
| Sociobiological Roots of Gaming, Ron Edwards | |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 3 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | 2 |
Why would a voter leave certain options unmarked? Because, when a voter ranks a product he is basically saying, "If my other choices all lose, give me vote to this one." Thus, the ballot, above, could be translated as follows:
I want D&D: The First 30 Years to win; but if not enough people support it, I want Hatred to Ambivalence: 30 Years of Church-Gamer Relations to win; but if not enough people support it, I want Dragonlance: A Tale of Merciless Exploitation to win; but I don't want either Sociobiological Roots of Gaming or 3E D&D: The Windows 95 of Gaming to win at all.
Choices that the voter would feel comfortable with winning should be ranked. Choices the voter would not feel comfortable with winning should be left unranked.
How Does One Count an AV/IRV Ballot?
Counting AV/IRV ballots is simple and well-described in the above web resources. In the first "round," all ballots are assigned to the first choice candidate of each voter. Let's assume that the sample election breaks down as follows:
| Candidate | Votes | % |
| D&D: The First 30 Years, E. Gary Gygax | 361 | 29.41 |
| Sociobiological Roots of Gaming, Ron Edwards | 374 | 30.48 |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | 18 | 01.47 |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 119 | 09.70 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | 355 | 28.93 |
| Total: 1227 |
If no candidate receives a majority of the vote, the bottom candidate is eliminated and the votes of those supporting it are transferred to the voter's second choice candidate.
| Candidate | Votes | % |
| D&D: The First 30 Years, E. Gary Gygax | 361+2 | 29.58 |
| Sociobiological Roots of Gaming, Ron Edwards | 374+6 | 30.97 |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | 18 | 01.47 |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 119+1 | 09.78 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | 355+9 | 29.67 |
| Total: 1227 |
As shown above, the people who supported the Parker book generally favoured Ranger as their second choice but all candidates gained votes due to its elimination. However, all candidates are far from a majority, with only one at over 30%.
As a result, the next lowest-scoring candidate is eliminated.
| Candidate | Votes | % |
| D&D: The First 30 Years, E. Gary Gygax | 363+38 | 32.84 |
| Sociobiological Roots of Gaming, Ron Edwards | 380+38 | 34.23 |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 120 | 09.78 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | 364+3 | 32.92 |
| Total: 1221 |
As shown above, the people who supported the Hickman book were divided perfectly evenly amongst the remaining competitors, except for six voters whose ballots must have looked like either:
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | |
| Sociobiological Roots of Gaming, Ron Edwards | |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 1 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | |
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | |
| Sociobiological Roots of Gaming, Ron Edwards | |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | 2 |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 1 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | |
or
| Candidate | Rank |
| D&D: The First 30 Years, E. Gary Gygax | |
| Sociobiological Roots of Gaming, Ron Edwards | |
| 3E D&D: the Windows 95 of Gaming, Stuart Parker | 1 |
| Dragonlance: A Tale of Merciless Exploitation, Tracy Hickman | 2 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | |
Still, none of the three remaining candidates has a majority. The bottom candidate is then eliminated and his votes are transferred to the remaining two competitors:
| Candidate | Votes | % |
| D&D: The First 30 Years, E. Gary Gygax | 401 | 32.84 |
| Sociobiological Roots of Gaming, Ron Edwards | 418+100 | 42.60 |
| Hatred vs. Ambivalence: 30 Years of Church-Gamer Relations, Jeff Ranger | 402+296 | 57.40 |
| Total: 1216 |
Thus, the winner would be Hatred to Ambivalence with the support of over 57% of voters.
The Runner-Up
Although AV/IRV is typically used in elections with a single victor, it is pretty easy to see how a runner-up and even a second runner-up can be declared. The runner-up in an AV/IRV system is the losing candidate in the final round; and the third-place finisher in the penultimate round.
|
